Lab notebooks

Lab Notebook: Enhancing Antimicrobial Ability of Lactobacillus plantarum Using Deep Learning and Directed Evolution Technologies

Project Title

Strengthening the Antimicrobial Ability of Lactobacillus plantarum Using Deep Learning and Directed Evolution Technologies

Experiment 1: Deep Learning Model Development

Objective:
Develop a deep learning model based on BiLSTM networks for accurate prediction of bio-functional peptides (BioPeps) across 13 categories, with a focus on antimicrobial peptides (AMPs).

Materials and Methods: 1. Data Collection:
- Compiled a dataset of 17,572 peptide sequences from public databases (e.g., BioPepDB, SATPdb, CancerPPD). - Filtered sequences to ensure diversity and quality, removing redundancy and sequences with unusual amino acids.

Feature Extraction:
Employed feature extraction techniques including:
- Amino Acid Composition (AAC)
- Dipeptide Composition (DPC)
- BLOSUM62 embeddings
- Physicochemical properties from AAindex
Model Architecture:
Developed three BiLSTM models: BiLSTM, BiLSTM_Attention, and BiLSTM_Predictor.
Integrated multiple features and evaluated 2,076 feature combinations for optimal prediction.
Training and Validation:
Split data into training (80%) and testing (20%) sets.
Used Adam optimizer with a learning rate of 0.001.
Evaluated model performance using Accuracy, Precision, Sensitivity, and AUROC.

Results:
- Achieved 94.8% accuracy in blinded testing and 88% accuracy in experimental validation for AMP prediction. - Best performance obtained with BiLSTM_Predictor for AMP, AHP, and DeP categories.

Conclusion:
The developed BiLSTM-based model significantly outperformed existing models in predicting BioPeps, demonstrating strong potential for practical applications.

Experiment 2: Dataset Collection and Processing

Objective:
Curate a comprehensive dataset for BioPeps prediction and refine data quality for improved model training.

Materials and Methods: 1. Data Sources:
- Collected sequences from databases like BioDADPep, CancerPPD, and AHTPDB. - Integrated high-throughput AMP-screening data.

Data Curation:
Removed sequences with over 90% identity to reduce redundancy.
Excluded peptides containing non-standard amino acids (e.g., B, J, O, U, X, Z).
Negative Control Selection:
Selected negative control peptides from UniProt, ensuring similar length distribution to BioPeps.

Results:
- Final dataset: 17,553 peptides, including 13 BioPeps categories and negative controls. - Successfully balanced dataset to reduce bias in model training.

Conclusion:
The curated dataset provides a robust foundation for training the deep learning model, ensuring high-quality and diverse peptide sequences.

Experiment 3: Cell-Free AMP Synthesis Platform Development

Objective:
Develop a cell-free synthesis platform for rapid and scalable production of AMPs.

Materials and Methods: 1. System Setup:
- Utilized E. coli cell extract-based protein expression system. - Prepared a 50 μL reaction mixture containing: - 50 mM HEPES buffer (pH 7.2) - 1.2 mM ATP, 0.85 mM UTP, GTP, CTP - 34 μg/mL folinic acid, 170 μg/mL E. coli tRNA - 5 μg/mL T7 RNA Polymerase, 15 μg/mL PCR product - 2 mM for each of the 20 amino acids

Reaction Conditions:
Incubated at 37°C for 4-6 hours.
Monitored AMP synthesis using SDS-PAGE and BCA Protein Assay.
Yield Optimization:
Tested various reaction volumes and component concentrations.
Achieved optimal yields of 0.5-2.1 g/L.

Results:
- Successfully produced AMPs in a cell-free system with high yields. - AMPs exhibited strong antimicrobial activity against target pathogens.

Conclusion:
The cell-free synthesis platform provides a scalable and efficient method for producing AMPs, suitable for industrial applications.

Experiment 4: Directed Evolution of Lactobacillus plantarum

Objective:
Enhance the antimicrobial properties of L. plantarum through directed evolution and high-throughput screening.

Materials and Methods: 1. Mutagenesis:
- Applied atmospheric and room temperature plasma (ARTP) mutagenesis to generate a diverse mutant library of L. plantarum.

Screening:
Used fluorescence-activated droplet sorting (FADS) to isolate mutants with elevated AMP production.
Designed a fluorescence-based sensor system with GFP10-E. coli and GFP11-B. subtilis strains.
Validation:
Tested isolated mutants against a panel of 7 pathogenic bacteria and fungi.
Measured antimicrobial activity and organic acid production.

Results:
- Isolated three mutant strains (M1, M2, M3) with significantly enhanced antimicrobial activity. - Mutants exhibited up to 2-fold higher bacteriostatic effects compared to the wild-type strain.

Conclusion:
Directed evolution combined with FADS effectively improved L. plantarum strains, making them suitable candidates for industrial antimicrobial applications.

Experiment 5: Industrial Application Testing

Objective:
Evaluate the effectiveness of enhanced L. plantarum strains in food preservation and pharmaceutical applications.

Materials and Methods: 1. Food Preservation:
- Tested in soy sauce, coconut juice, bread, and lean meats. - Monitored microbial growth and product quality over time.

Pharmaceutical Testing:
Assayed cell-free synthesized AMPs against multi-drug resistant bacteria.
Conducted minimum inhibitory concentration (MIC) tests.

Results:
- Enhanced L. plantarum strains effectively inhibited spoilage and pathogenic microorganisms in various food matrices. - Cell-free synthesized AMPs showed potent activity against drug-resistant bacteria.

Conclusion:
The enhanced strains and cell-free AMPs have broad applications in food and pharmaceutical industries, offering natural alternatives to synthetic preservatives and antibiotics.

Notes and Observations

Continuous optimization of model parameters improved prediction accuracy.
ARTP mutagenesis was crucial for generating diverse mutants.
The combination of computational predictions and experimental validation significantly reduced time and cost in identifying effective AMPs.

Future Directions

Expand the dataset to include more diverse BioPeps sequences.
Further optimize the cell-free synthesis platform for other peptide categories.
Conduct large-scale validation of L. plantarum strains in real-world industrial settings.